Question Answering on NQ (F1 and EM)
43.62F1PRAISE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRAISEApproach=Ours2026.04 | 43.62 | 22.6 | |
| R1-SearcherApproach=Agentic Search2026.04 | 41.15 | 21.61 | |
| Search-R1Approach=Agentic Search2026.04 | 40.98 | 20.78 | |
| TIPSApproach=Process Reward2026.04 | 40.38 | 20.3 | |
| StepSearchApproach=Process Reward2026.04 | 39.79 | 19.81 | |
| ReasonRAGApproach=Process Reward2026.04 | 39.65 | 19.78 | |
| Naive RAGApproach=Standard2026.04 | 37.26 | 18.59 | |
| Search-o1Approach=Agentic Search2026.04 | 32.43 | 15.71 | |
| Direct LLMApproach=Standard2026.04 | 17.16 | 5.98 |